Mashable News Article Popularity Prediction for AutoML Benchmarking

A challenging tabular dataset for predicting the log-scaled popularity of Mashable.com news articles based on title text and auxiliary numerical features. The dataset, sourced from a 2021 arXiv paper, is intended as a difficult benchmark for AutoML systems. Authors include Xingjian Shi, Jonas Mueller, Nick Erickson, Mu Li, and Alexander J. Smola.

Use Cases

• Benchmarking AutoML performance on difficult prediction problems based on the description.
• Predicting article popularity on social media based on title text features.
• Analyzing the relationship between article metadata and virality based on features like word count and keyword number.
• Developing models for weak signal prediction based on crude numerical summary statistics.

Strengths

• Designed as a challenging benchmark for AutoML systems, as stated in the description.
• Derived from a real-world source (Mashable.com) and associated with a published arXiv preprint.
• Includes a mix of text-derived and numerical features for prediction.

Limitations

• Row count, column details, and sample data are unknown, limiting suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.
• The description notes the prediction problem is fundamentally hard with only weak signal from features.

Provenance

Source

Mashable.com, via the paper 'Benchmarking multimodal automl for tabular data with text fields' (arXiv:2111.02705).

Collection Method

Likely collected via web scraping or API access to Mashable.com articles.

Time Range

null

Freshness

Last update date is unknown; freshness unverified.

Geography

null